[aarch64-port-dev ] Freeze aarch64/jdk8

[Andrew Dinn]

On 25/02/16 13:38, Andrew Haley wrote:
 . . .

>> 8136596.patch
>> revid: 8561
>> 8136596: Remove MemBarRelease when final field's allocation is NoEscape
>> or ArgEscape
> 
> This is a minor optimization, not submitted to jdk8u.  It's also
> somewhat risky.  OK if you make it AARCH64_ONLY.

Ok, will do.

>> volcas.patch
>> revid: 8568
>> Backport optimization of volatile puts/gets and CAS to use ldar/stlr
> 
> It's probably safe, but there have been significant reworkings in
> this area.
> 
> Maybe make this AARCH64_ONLY ?  Please have a look at the generated
> code when running G1.

This shared change reverts part of a modification you had previously
made to the shared code in an earlier attempt to implement optimization
of volatile puts on AArch64. So, applying this patch merely restores the
status quo as regards the shared code that is currently in both jdk8 and
upstream jdk8u.

Also, the same reversion was applied and is still present in jdk9. It
has no bearing on any of the reworkings that have happened since the
original patch was added to jdk9 and I don't envisage it having any such
effect.

So. I don't think there is any reason to make this AARCH64_ONLY.

>> 8138966.patch
>> revid: 8573
>> 8138966: Intermittent SEGV running ParallelGC
> 
> Yes, serious crasher bug.  This is already in jdk8u.
> http://hg.openjdk.java.net/jdk8u/jdk8u/hotspot/rev/110735ab93ec
> 
> Please Ask Andrew Hughes about this one: it should already be in.

We already agreed that I would include this as part of my patch.

>> largecodecache.patch
>> revid: 8576
>> Add support for large code cache
>>
>> this makes changes to two files.
> 
> The non-AArch64-specific parts of this patch are not used by anything
> so should not be included.  The rest is OK.

Ok, I will rework this to include only the AArch64-specific code.

>> 8145438.patch
>> revid: 8583
>> 8145438: Guarantee failures since 8144028: Use AArch64 bit-test
>> instructions in C2
>>
>> this makes a small change to src/share/vm/adlc/formssel.cpp
> 
> This is OK.  It can't be submitted to jdk8u upstream because
> it's only needed for Arch64.  make it AARCH64_ONLY, just for
> safety.

ok, will do.

I'll build and test with the revised patches and report (this tim ewith
with a webrev) when I have managed to get it to pass basic smoke tests.

regards,


Andrew Dinn
-----------